Neuron Explorer

Qwen3-0.6B

"be" after a modal verb

A neuron in layer 19 of Qwen3-0.6B. Its strongest snippets are very consistent: they mostly peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the word "be" when it follows a helper word like "can" or "can also", as in "can be used". When turned down, the model's text shifted toward protection and safety topics, and when turned up it wandered on without a clear theme.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. Think it is wrong? Flag it (opens a prefilled issue on GitHub).

Fires most on

The text windows where this neuron's activation was highest. Amber marks positive activation; the outlined token is the peak. Hover a token for its strength.

  1. #1 peak

    to healthy plants during the first week for flowering.⏎Other Uses For PK Booster⏎PK boosters can be used not only to help promote healthy flowering but can also be used if your plants have a PK deficiency you should apply

  2. #2 peak

    every second. But as traditional emotion metrics can always be improved, new classification methods be tried out. In a recent IEEE paper, we demonstrate how machine learning can be used to identify the two main dimensions of emotions in EEG signals: arousal and val

  3. #3 peak

    microscope. A related method is called “rational drug design,” where scientists “build” drug molecules specifically so they can fight HIV in targeted ways. Computers can be used to simulate a chemical compound and to design chemical structures that might work against HIV

  4. #4 peak

    mechanical properties can be determined during the acquisition of an image and subsequently used as an indirect chemical probe. Beyond analysis under ambient conditions, atomic force microscopes can easily be outfitted for detailed relative humidity and reactive gas studies to further probe processes of chemical

  5. #5 peak

    arguing. (Here, defuse is being used figuratively.)⏎DiffuseThe adjective diffuse means spread out or not concentrated in one place. Diffuse can also be a verb meaning to circulate or to spread.⏎- She is the CEO of

  6. #6 peak

    it with a computer controlled laser The process works in air at room temperature and eliminates the ... - Read More⏎UCL scientists have shown how advanced computer simulations can be used to design new

Steering

Push this neuron up or down while the model writes, and watch what changes. The outputs were generated ahead of time, so nothing runs in your browser.

turned down a lot
Prompt

Here is something I have been thinking about lately: I have been told that the range of a model that I was once studied with my neighbor was estimated to be covered by the range of the model I was once studied with my neighbor, and I have been told to estimate the range of the

Most negative on

Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.

  1. #1 peak

    again. The Cyberknife technology works so precisely that the administered dose only reaches the tumor and spares surrounding tissue, organs or nerves.⏎Which tumors and lesions can be treated with Cyberknife?⏎The Cyberknife System is designed for radiosurgical treatment of

  2. #2 peak

    factors that affect the objective function should be written in the form of constraints.⏎The constraints and the objective are linear functions.⏎Only problems that satisfy the above requirements can be solved by using the linear programming models. Once linear programming is found to be applicable

  3. #3 peak

    When the soil has been properly prepared, spread⏎the seed at the appropriate time and rate⏎for your species.⏎to spread seed⏎- Measure the area to be planted and determine the total amount of seed⏎-

  4. #4 peak

    an index to materials at all levels of<|endoftext|>Find a Radiation Oncology Specialist, at Home or Abroad⏎Radiation Oncology Specialist Doctor⏎A disease can be detected not only through laboratory or pathology tests, but also through the use of various

Connections

Show on map

Which neurons in earlier layers feed this one, and which later ones it feeds. Amber links excite, blue links suppress; longer bars are stronger.

Fed by earlier layers

  1. L18:960 strongly suppresses
  2. L18:1744 strongly suppresses
  3. L18:2476 excites
  4. L17:1755 suppresses
  5. L18:727 suppresses
  6. L18:1930 suppresses
  7. L18:2703 suppresses
  8. L18:1243 suppresses
  9. L16:2681 excites
  10. L17:2353 excites
  11. L14:1887 suppresses
  12. L16:2912 suppresses
  13. L10:1736 suppresses
  14. L18:1353 suppresses "the" after "and" in academic prose
  15. L9:877 excites
  16. L17:1012 excites
  17. L15:599 excites
  18. L10:1364 suppresses
  19. L5:453 excites
  20. L17:2575 suppresses
  21. L18:2002 suppresses
  22. L7:800 excites
  23. L4:116 suppresses
  24. L15:1210 weakly excites

Feeds later layers

  1. L20:330 suppresses
  2. L21:656 suppresses
  3. L22:1170 suppresses
  4. L20:1872 suppresses
  5. L21:1293 suppresses
  6. L20:1422 suppresses
  7. L24:1984 suppresses
  8. L20:480 weakly suppresses
  9. L24:2319 weakly suppresses
  10. L24:1422 weakly suppresses
  11. L25:899 weakly excites
  12. L25:1549 weakly suppresses
  13. L21:1606 weakly suppresses
  14. L20:2940 weakly suppresses
  15. L20:254 weakly suppresses
  16. L23:1795 weakly suppresses
  17. L21:1610 weakly suppresses
  18. L20:922 weakly suppresses
  19. L26:135 weakly excites
  20. L21:1131 weakly excites
  21. L27:1601 weakly suppresses
  22. L20:1623 weakly suppresses
  23. L20:1726 weakly suppresses
  24. L23:2410 weakly suppresses

Activation histogram

How often this neuron fires at each strength. Almost always it does close to nothing (the tall bars in the middle); the snippets above are the rare moments it fired hard, amber for positive and blue for negative.

1 values between -17.66 and -16.971 values between -14.92 and -14.241 values between -12.19 and -11.511 values between -8.78 and -8.091 values between -7.41 and -6.731 values between -6.73 and -6.052 values between -6.05 and -5.365 values between -5.36 and -4.682 values between -4.68 and -4.008 values between -4.00 and -3.313 values between -3.31 and -2.6310 values between -2.63 and -1.9522 values between -1.95 and -1.26167 values between -1.26 and -0.582686 values between -0.58 and 0.10917 values between 0.10 and 0.7939 values between 0.79 and 1.478 values between 1.47 and 2.1510 values between 2.15 and 2.833 values between 2.83 and 3.521 values between 3.52 and 4.204 values between 4.20 and 4.883 values between 4.88 and 5.575 values between 5.57 and 6.251 values between 6.25 and 6.933 values between 6.93 and 7.621 values between 7.62 and 8.301 values between 8.30 and 8.981 values between 8.98 and 9.66offnegativepositiveoften

Logit lens

The words this neuron nudges the model toward (amber) or away from (blue) when it fires. A hint, not a label: many neurons work through later layers instead.

Promotes

  1. rael
  2. .Reader
  3. �能
  4. .NewReader
  5. /></
  6. cbd
  7. IFORM
  8. eselect
  9. igin
  10. ubbles

Suppresses

  1. covered
  2. 处理
  3. treatment
  4. treated
  5. tratt
  6. Handling
  7. Treatment
  8. 保护
  9. coverage
  10. Coverage